A test of whether selection maintains isochores using sites polymorphic for Alu and L1 element insertions.

ONE of the most striking features of the human Although we have little direct information about the pattern of element integration, we can use sites that are genome is the large-scale variation in GC content that is found along chromosomes. This variation was polymorphic for SINE and LINE insertions to investigate whether selection on composition affects the distridiscovered in 1973 (Filipski et al. 1973) and was originally thought to be organized as a series of “isochores,” bution of elements, since sites that are polymorphic for element insertions will reflect the pattern of integration blocks of DNA of homogeneous base composition separated by borders of sharp transition (Bernardi 1995). more closely than sites at which the element is fixed. We therefore tested whether the sequences flanking While more recent studies of long sequences (Nekrutenko and Li 2000) and the draft human genome (Insites polymorphic for Alu and L1 elements are different in composition from the sequences flanking elements ternational Human Genome Sequencing Consortium 2001) have suggested that the mammalian genome fixed within the human population. Since the distribution of fixed elements may also be affected by the rate probably does not fit the isochore model exactly, it is still evident that base composition varies over scales as at which elements decay, we have restricted our comparisons between fixed and polymorphic elements to elelarge as 10 Mb. The origins of the large-scale variation in GC content ments of the same subfamilies. Mark Batzer, Prescott Deininger, and colleagues have remain controversial (Eyre-Walker and Hurst 2001). It has been suggested that it could be a consequence of investigated whether a number of Alu insertions from young Alu subfamilies are polymorphic in humans (i) mutation bias (Filipski 1987; Suoeka 1988; Wolfe et al. 1989), (ii) natural selection (Bernardi and Ber(Batzer et al. 1995, 1996; Arcot et al. 1996, 1998; Sherry et al. 1997; Roy et al. 1999; Roy-Engel et al. nardi 1986; Hughes and Yeager 1997; Eyre-Walker 1999), or (iii) biased gene conversion (Holmquist 2001). Using their primer sequences and a BLAST 1992; Eyre-Walker 1993). In this study we test whether search of the draft human genome, we located the genonatural selection is acting upon base composition by mic position of 240 of their elements: 86 Ya5, 12 Ya8, considering the distribution of SINE and LINE ele11 Yb8, 34 Yb9, 90 Yc1, and 7 Yc2; 76 of these sites ments. were polymorphic for the element. Similarly we used If selection is acting upon the base composition of the primer sequences or accession numbers given by isochores, then the distribution of newly integrated eleBoissinot et al. (2000) to locate 41 young Ta L1 repeats: ments, and elements that are fixed within the popula14 Ta-0 and 27 Ta-1; 18 of these sites were polymorphic. tion, should be different. For example, if selection favors All the sites polymorphic for an element appear to be high GC content in certain regions of the genome, a consequence of insertion rather than deletion. From then one might expect the GC content flanking newly these data we extracted the 500 and 5000 bp each side inserted L1 elements, which are still polymorphic in the of the repeat, or the site midway between the two primpopulation, to be greater than the GC content flanking ers, if the element was missing from the genomic seyoung L1 elements that are fixed within the species. quence. All the data are published as supplementary This is because L1 elements are GC poor; they will information. For comparison we also extracted 200 rantherefore be selected against if they integrate into the domly chosen sequences of 1 and 10 kb from the human GC-rich parts of the genome. genome. Figure 1 shows the distribution of the GC content of the 10 kb of sequence flanking sites polymorphic or 1Corresponding author: School of Biological Sciences, University of fixed for Alu or L1 insertion, along with the distribution Sussex, Brighton BN1 9QG, United Kingdom. E-mail: [email protected] of randomly selected sequences of the same length;